OpenAI tests Zero-Retention safety system for frontier models

OpenAI introduced Private Safety Processing, a system detecting misuse patterns across AI interactions without employee access to prompts or responses, preserving Zero Data Retention for eligible API customers. Testing with early customers begins now, with rollout and a technical white paper planned for September.